
A stereo pair should not be aligned as though its two photographs were interchangeable samples for averaging. Hugin's align_image_stack -S permits horizontal displacement between corresponding points. That retained relationship is part of stereoscopic presentation, while accidental vertical disagreement needs review. The tool also offers stereo-window choices; it does not certify that a particular pair will be comfortable in every viewer.
Preserve the left/right relationship
Start with the authorized left and right photographs and keep their identities explicit. Do not swap them simply because a file list sorts differently. Inspect corresponding foreground and background details before processing.
This differs from repairing a panorama seam. The panorama disagreement guide explains why viewpoint differences prevent all depths from coinciding. In stereo work, some horizontal difference is intentional rather than evidence that every point should be forced together.
Use the stereo alignment contract
The official align_image_stack reference documents -S as allowing horizontal control-point shift and creating horizontal-line constraints. For an illustrative pair:
align_image_stack --use-given-order -S -p stereo-check.pto -a stereo left.tif right.tif
The explicit order option avoids the documented default exposure-value sorting. -p retains a project for inspection; -a requests remapped TIFFs. These are proposed filenames and a suggested invocation, not an executed rectification result.
Do not replace this with a focus-stack averaging command. Matching a shared point's complete position and retaining stereo disparity answer different requirements.
Decide where the stereo window belongs
The same manual provides two additional choices, each enabling stereo mode:
| Option | Documented reference decision |
|---|---|
-A | Uses the nearest point for a window intended to place the scene behind the display plane |
-P | Uses a nearest point in border regions, permitting central content to appear in front |
Select the rule according to the actual presentation and inspect the result. The descriptions do not guarantee that an arbitrary source pair has an acceptable disparity range or that a central subject always produces a successful pop-out effect.
Keep this separate from physical measurements. The planar-versus-lens-distortion check illustrates another reason one simple alignment model may not fit all recorded geometry.
Review the delivered pair as stereo
In the intended viewing arrangement, inspect corresponding features near the center, edges and different depths. Check left/right identity, vertical mismatch, framing and whether foreground content meets the chosen window sensibly.
If the pair cannot be viewed acceptably, stop rather than approving two individually sharp TIFFs. Keep the original pair and project so the alignment can be investigated. No stereo pair, display comfort evaluation or geometric measurement was performed for this guide.
Related tasks
- How Do I Repair a Panorama Stitch Without Duplicating a Building?
- Why Does a Screen Replacement Fit the Corners but Bend Away in the Middle?
Sources and scope
- Hugin align_image_stack: Generating stereo image pairs: -S permits horizontal shift and uses horizontal-line control points; -A and -P add distinct stereo-window reference choices and imply -S; -p saves project and -a remapped TIFFs.
Research checked on 2026-10-23 (Asia/Shanghai) against actually opened official manual/RawPedia pages. No camera capture, command execution, raw processing, calibration or display test was performed. Examples and acceptance checks are proposed, and the cover is decorative.
Prepared by Imagild Editorial, the company editorial team responsible for these guides. AI assisted research and drafting. This is documentation-based guidance with proposed examples, not a claim of hands-on testing or an independent product review. Report a correction.
Cover: Mariya Eskina / Pexels, Pexels License. This is an illustrative source photograph, not an application screenshot or a generated test result.